Without exception the companies that weighted academic algorithm skills over practical architecture skill and implementation experience had some of the most confident and brazen mistakes in security or scalability and some of the poorest work-life balance.
I understand most people will say "this is too much work". The problem is that it is all too easy for someone to say "in my experience, X is correlated with Y" ... but that isn't very believable if the observer didn't even write things down.
I don't like asking for everything to be quantified, but it does seem in this case you probably have the data in your head and just need a nudge to write it down.
But it was his comment you chose to "suggest he show the work".
Also, I didn't say it didn't apply. I'm not a perfectly consistent machine you know.
The industry is at a point where there are plenty of clueless cookie cutter companies cargo culting Leetcode just because those interesting companies engage in it.
As far as interesting companies that don't require that, there's at least Stripe, famously, and surely more from this list:
> Programming/debugging phone screen + on site with your own laptop/setup and full access to internet, systems design discussion and talk with hiring manager about team alignment.
That seems to me to be a technical interview. I'm not sure what makes this different from a Facebook or Google.
I have quite high standards when I hire, which is why I don't use leetcode to assess. :)
I can get much better data more quickly without it. And, indeed, the best places I've ever worked with the smartest people I've ever known have not used these types of problems in the interview.
I've personally solved hundreds of them for the problem-solving challenge, but it's a horrible way to determine if someone is a good fit for a job. It's like interviewing a mechanic for your auto shop by seeing how well they can change a tire using only a screwdriver.
How would you go about checking if somebody is qualified in an interview?
Using his past experience may help but too often it can be embellished or flat out lied about. That's why these algorithm interviews became popular in the first place, to my knowledge.
It depends on the job, but for coding I've had great luck asking them to bring some code they were proud of in to the interview. Language doesn't matter, subject doesn't matter.
And then in the interview I ask them to teach me how it works. It becomes apparent in no time if they don't know what they're doing (or stole it). And if they can tell me about it, I learn how well they understand their own system--do they only understand it well enough to code it, or do they understand it well enough to teach it? I also learn if they're a good communicator. And if they're a good culture fit. And if they're enthusiastic about coding. I can also ask probing questions about design decisions and shortcomings.
That covers the "how well do they code" part, but falls a little short on the problem-solving part. But I can come up with on-the-fly questions (often about the code they brought in) that exercise those muscles. "What would you do if this data weren't available?" "What if you needed more guaranteed uptime?" "What if you needed to process 1000x more data in the same timeframe?" These are more relevant questions.
I'm after someone who learns fast. I'll take someone who knows a little and learns fast over someone who knows a lot and learns slow 9/10 times.
(Once I had someone bring in a device driver written in C for a JavaScript gig. They were hired. Another time we hired a dev with zero experience in the platform, language, or framework. Worked out great.)
If they ace leetcode-style challenges, I know they're good at _that_, but I don't trust it as a proxy for the other things I want to know, above.
The parent commenter wasn't saying we shouldn't have "high standards"; just that filtering for maximal prowess at leet code grinding does not serve as a useful instance of such, in their book.
Further, this distinction seems basically quite obvious -- there's no other way to read their text, actually.
Other industries don’t hire this way for anyone but fresh grads, and most don’t even do this for them.
Companies that have high standards are far more interesting places to work at and with more interesting colleagues. They all require whiteboards and leet code.
It's definitely true that not all companies that use these algorithm interviews are worth working for. But, at least in my experience, companies that don't require them at all are pretty awful places to work. It demonstrates that management doesn't care who they hire.